Date of Award

2000

Degree Name

Physical Science

College

College of Science

Type of Degree

M.S.

Document Type

Thesis

First Advisor

James Brumfield

Second Advisor

Ralph Oberly

Third Advisor

Dewey D. Sanderson

Abstract

Delineation of wetlands is an important aspect of protecting of these ecologically diverse and biocomplex habitats. Selected image-processing techniques are applied to remotely sensed imagery of Canaan Valley, in order to determine the best method of extracting wetland features for habitat delineation. Remotely sensed data from April 1997 and September 1999 were compared to determine the accuracy of feature extraction, and the effect of seasonal and long term climate change on wetland delineation. Pattern recognition for feature extraction was used to identify vegetation associations and changes in habitats. Computer simulation of the infrared band was performed using ER Mapper so as to improve radiometric techniques. Natural clusters were identified using a combination of PCA and clustering algorithms in Idrisi and ER Mapper. The clusters reflect common spectral reflectance of pixels within the image; these are classified as vegetation type based on GPS sampled ground truth data. Data conversion between a combination of image processing and GIS software (ArcView, ER Mapper and Idrisi) was used to improve feature extraction. The selectcombination of image processing techniques is important in extracting relevant features. Contingency tables and photointerpretation is used to determine the accuracy of feature extraction with ground truth data. A geobiophysical model, consisting of spatially georeferenced layers of digital imagery, vegetation communities, microtopography and hydric soil zones illustrates the habitat diversity associated with wetlands. The inclusion of microtopography or a hydric soil zone map can improve feature extraction of swampy vegetation and areas of low drainage. Plants associated with hydric soils and areas of drainage are clearly distinguished from larger shrubs and trees on the classified aerial imagery. Vegetation types and their associated environment are correlated with previous field research in the Canaan Valley area.

Subject(s)

Imaging systems - Image quality.

Habitat (Ecology)

Wetland ecology.

Wetlands - West Virginia.

Canaan Valley, (W. Va.)

Download

Share

COinS